Cape Town - 2026 ISMRM-ISMRT Annual Meeting and Exhibition - AI To Make Protocols, Plan, QC, and Correct Motion

AI To Make Protocols, Plan, QC, and Correct Motion

Digital Poster

Acquisition & Reconstruction

Tuesday, 12 May 2026

Digital Posters Row B

09:15 - 10:10

Session Number: 461-02

No CME/CE Credit

Similar Sessions

This session shows AI methods used to plan the protocol, to plan the scan, to QC the data or to correct motion.

		461-02-001. Patient-tailored Protocol Prescription using Physics-informed Large Language Model Agents Anuj Sharma, William Grissom, Mark Griswold Case Western Reserve University, Cleveland, United States of America Impact: Selection of clinical protocols is prone to human errors which contribute to systemic loss of $1.04 billion/year. We demonstrate a physics-informed LLM system that can query the EHR-database and prescribe protocols with statistically maximum information in less than 30 seconds/patient.
		461-02-002. Toward Intelligent MRI Workflows: Evaluation of AI-Driven Protocol Selection and Automated Workflow Triage Siddhant Dogra, Jordan Kondo, Julian Wohlers, Rainer Schneider, Jens Gühring, Fabian Wagner NYU Langone Health, New York, United States of America Impact: Large language models can transform MRI operations by interpreting clinical history to automatically select brain MRI protocols and distinguish cases suitable for automated versus standard human-driven workflows, reducing manual burden and improving scanner throughput.
		461-02-003. Automated landmark detection in four-chamber cardiac MRI using deep learning for advanced view prescription Gaspar Delso, Collin Buelo, Justin Leonard, Fara Nikbeh, Jane Names, Willian Cordeiro, Martin Janich GE HealthCare (ES), Spain Impact: This study advances automated cardiac MRI by enabling reliable landmark detection in 4CH views, streamlining right ventricle view prescription and improving workflow efficiency.
		461-02-004. Manual vs automated planning of cardiac MRI planes: A reproducibility study across different field strengths Margarita Gorodezky, Gaspar Delso, Karolin Deyerberg, Lena-Maria Watzke, Ann-Christin Klemenz, Mathias Manzke, Antonia Dalmer, Roberto Lorbeer, Marc-André Weber, Benjamin Böttcher, Felix Meinel GE Healthcare, Munich, Germany Impact: Automated cardiac planning ensures high-quality, reproducible plane prescription across magnetic field strengths in healthy volunteers. It standardizes workflow, minimizes operator variability, and reduces dependence on radiographer expertise, enabling consistent imaging and allowing clinicians to prioritize patient care.
		461-02-005. SurfScribe: Cortical surface-driven automated online slice prescription applied to ultra-high-resolution vascular MRI Karthik Gopinath, Paul wighton, Mukund Balasubramanian, Robert Frost, Renzo Huber, Kyle Droppa, Douglas Greve, Jesper Nielsen, Oula Puonti, Daniel Haenelt, Andre van der Kouwe, Bruce Fischl, Jonathan Polimeni, Juan Iglesias, Divya Varadarajan Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Charlestown, United States of America Impact: SurfScribe enables online, anatomy-informed MRI slice or slab prescription using cerebral cortical surfaces generated within minutes at the scanner. It supports motion-robust, vessel-targeted imaging and unlocks new possibilities for automated scan planning in high-resolution neuroimaging research and clinical workflows.
		461-02-006. Feasibility of Inline Deep-Learning–Enabled Gold Fiducial Detection on MRI Using OpenRecon for MRI-Only Prostate Radiotherapy Jonathan Goodwin, Ashley Stewart, Matthew Richardson, Steffen Bollmann Calvary Mater Newcastle Hospital, Newcastle, Australia Impact: Accurate identification of gold fiducial markers is essential for MRI-only prostate radiotherapy planning. Integrating deep-learning enabled fiducial detection into MRI reconstruction using OpenRecon provides a practical imaging pipeline that facilitates clinical implementation of MRI-only prostate treatment planning for radiotherapy
		461-02-007. Real-time MRI-based fetal femur length measurement Johannes Barcsay, Sara Neves Silva, Jordina Aviles Verdera, Charline Bradshaw, MARY RUTHERFORD, Susanne Schulz-Heise, Jana Hutter Technische Hochschule Deggendorf, Germany Impact: Fully automatic planning, acquisition and measurement of fetal femur length in-utero on MRI, developed and tested retrospectively and prospectively in 72 and 24 cases, enables detailed individualised growth assessment and earlier detection and prediction of pathologies.
		461-02-008. Needle-Free Myocardial Blood Flow and Reserve Quantification Using AI-Enhanced Coronary Sinus Flow MRI Manuel Morales, Alexander Schulz, Nicole C.Y. Deng, Tess Wallace, Warren Manning, Reza Nezafat Beth Israel Deaconess Medical Center and Harvard Medical School, Boston, United States of America Impact: We evaluated feasibility of AI-enhanced phase-contrast CMR for quantifying coronary sinus flow during post-exercise, demonstrating excellent repeatability and good correlation with quantitative myocardial perfusion. Our technique potentially enables accurate, contrast-free, pharmacology-free assessment of myocardial blood flow and coronary flow reserve.
		461-02-009. Rigid Head Motion Estimation from Ultra-Short Lissajous Navigators at 7T using Geometric Deep Learning Stanislav Motyka, Bernhard Strasser, Lukas Hingerl, Aaron Osburg, Hauke Fischer, Georg Langs, Wolfgang Bogner Medical University of Vienna, Vienna, Austria Impact: This work enables fast, data-driven motion estimation directly from ultra-short MR navigators at 7T, supporting the development of real-time motion correction methods and improving the reliability of high-resolution neuroimaging at ultra-high field strengths.
		461-02-010. A Self-Supervised Transformer for Myelin Water Fraction Mapping from Multi-Echo Gradient-Echo Imaging Samiha Prima, Yang Gao, Hongfu Sun The University of Queensland, Brisbane, Australia Impact: The proposed model leverages cross-attention to capture temporal dependencies across echoes, enabling robust myelin-water-fraction estimation under diverse noise and acquisition settings. A self-supervised training paradigm eliminates dependence on in-vivo dependence, and reconstruction is considerably faster than conventional nonlinear least-squares fitting.
		461-02-011. Robust quantification of CBF and ATT in multi-delay PCASL with fewer PLDs and averages using a CNN-Transformer framework Shibiao Wang, Yining He, Sang Hun Chung, Lirong Yan Northwestern University, Chicago, United States of America Impact: A CNN-Transformer model is proposed for robust CBF and ATT quantification from multi-delay pCASL with reduced number of repetitions, maintaining high accuracy and noise resilience. It also enables optimizing PLD sampling strategies under fixed repetitions for efficient clinical acquisition design.
		461-02-012. Deep blind arterial input function correction in perfusion: Resolving zero outputs through sigmoid activation Habib Rebbah, Corentin Cluet, Timothe Boutelier Olea Medical, La Ciotat, France Impact: This work eliminates unexplained artifacts in deep blind AIF correction, improving reliability for quantitative perfusion MRI. It enables stable prediction for higher temporal resolutions and supports safer translation of deep-learning–based AIF estimation into clinical workflows.
		461-02-013. Population-prior-assisted Implicit Neural MRI Reconstruction for Improved Generalization Across Undersampling Patterns Chushu Shen, Hengjie Liu, Dan Ruan, Debiao Li Cedars-Sinai Medical Center, Los Angeles, United States of America Impact: Our framework enables sampling-pattern-agnostic reconstruction for accelerated MRI. It is simple, effective, and robust. The highly flexible framework also motivates more advanced prior design for further improvements and clinical adaptation.
		461-02-014. MeanFlow Perceptual Loss (MFPL) for Low-Field MR Image Enhancement via Complementary VAE and SiT Features Zechen Zhou, Long Wang, Ajit Shankaranarayanan Subtle Medical Inc, Menlo Park, United States of America Impact: The MeanFlow Perceptual Loss (MFPL) method uniquely combines VAE and SiT features, yielding state-of-the-art perceptual quality and structural fidelity. It demonstrates robust generalization across different low-field scanners, accelerating the clinical adoption of clearer, more reliable images from portable MRI devices.
		461-02-015. Image quality assessment of MR image enhancement with large vision-language models Caohui Duan, Dong Zhang, Jianxing Hu, Xiaonan Xu, Youmin Li, Xin Lou The First Medical Center, Chinese PLA General Hospital, Beijing, China Impact: Our study demonstrates the potential of large vision-language models in MR image quality assessment, suggesting their capability in supporting or even replacing radiologists in evaluating MR image enhancement and other clinical scenarios, such as scan quality control.
		461-02-016. Adaptive 3D Vision-Based Multimodal Physiological Gating System for MR Imaging SAGNIK GHOSH, SANKET MALI, HARIKRISHNA RAI GE HealthCare (Bengaluru, India), Bengaluru, India Impact: This contactless 3D-vision based AI-driven gating solution transforms MR imaging by eliminating hardware-sensors, improving patient comfort, reducing setup-time, and enabling real-time motion-free acquisition—empowering radiologists with adaptive physiological gating, sharper images, higher throughput, and new research opportunities in personalized, precision MR-imaging.